Lacks a little presence on the profile, if you can ad it, it's the same. But the question it's : how does it feel when playing ? Do you feel a big difference in dynamics and sensitivity ?
Hey Ola - regarding the time to create the captures it is relative to what hardware you using to do it - your computer and it’s graphics card apparently - Jason Sadites swapped graphics cards on his machine to an nvidia specified and cut the capture times down to sub 10 minutes for the fast and normal and I believe 27 mins for the advanced super macho capture - I assume you were using your laptop - sounds like for faster captures you need a beefier computer rig. Probably makes sense if your doing a ton of captures for resale online or something.
I have a bunch of great amps in my home studio and got really excited when the new tone x came out with the ability to capture my own amps. The very first time I tried it was with my own radial reamp box out of my interface and I was modeling my Marshall slp 100 with a 4x12 Marshall cab filled with g12h celestions and a royer 121 mic. I went through the entire capture process and chose the "regular" training time. The end result sounded really good but it sounded like my plexi with a pillow over the speaker cab. I then purchased the tone x box itself and ran through the process again using the tone x box to do the re amping work and the box made it sound a touch better but again it lacked the high end clarity of my actual amp mic'ed up. I then starting experimenting around, ditched the ribbon mics and used an sm57 on my Marshall jcm900 and whammmmm. The final capture sounded absolutely identical to my mic'ed up amp. I mean identical and this was using the normal training time. I found the longest training time to only take 1 hour on my iMac but didnt noticed enough of a difference in the final captures over the 23 minutes the "regular" training time took. The key seems to be using a mic that isn't overly dark in the first place. When using the sm57, the average frequency response seems to allow the capture to process in a way that doesn't allow it to become to boomy or filled with low end. I went back and tried my plexi with the royer 121/sm57 combo and leaned the capture mix of the two mics towards the 57 and whammm, the plexi sounded spot on to my mic'ed up tracks going straight into the interface. I now am truly impressed with the tone x capturing. The very best results are making sure you have good high end on the amp settings itself. Meaning, my plexi, tends to be shrill in the high end if I don't back the treble off while playing but when capturing the amp, be sure to allow plenty of this high end through so it can capture the most honest all around tone for your final capture. Just my 2 cents to help others!
it's amazing the fact that +10 years after the kemper release we still comparing the tone cloning capabilities vs. him .I mean,10 years it's a lot in terms of technology advances
Ahhhm I don't think the kemper is nothing special in terms of technology, its brilliance is that it is efficient for the hardware and it established a different mode of working and sharing sound. However it has a lot of sound flaws IF you care so much about that. I mean some of the stuff that bug me from the kemper other types of players (metal vs fusion vs blues etc) would disregard, won't notice or even come in contact with. And in the mix is good enough for many top producers. But this IK multimedia absolutely wins on cost, performance and the ability to just fucking plug it in a daw.
I never liked the kemper. When working as FOH mixer, many guitar guys comming at concerts with that box and always sounding fake in a live situation. It was funny hearing them "but that was the vintage Marshloow model made by Merschuga's guitarrist... it can't be bad". LoL
I think its important to say, that there is a EQ section in the top left corner where you can set up the freqency range for bass, mid and treble. It helps a lot^^ And remember, mesas are really hard to capture. I think the advanced mode will give a bit more sparkle to the tone. On other systems these mode just need 1,5 hours. The Capture got the charackter of the real deal but perhaps a bit tooo much gain =P
@@robertmazurowski5974 huh interesting. Are you sure you aren't imagining this because of some sort of bias? I mean, would you be able to tell apart a capture and a real amp in a blind test? I don't think I would.
Being that accurate translating the rig tone I believe it is worth the wait on the processing time. Better to spend some extra time getting the perfect tone before and doing the job just once. Great review btw. Thanks!
In machine learning longer training time doesn't essentially mean a better model (over fitting, less generalisation). I wonder how exactly this was optimized here...
@@CraftingCake Hi. I understand it could be not proportionally better, but shouldn't it be better? You mean then that a longer training time could result in zero improvement or worse, a negative improvement?? I'm not into machine learning, but made my comment based purely on common sense
Great vid, you don"t know me but you mentioned my captures up for free on their Tonenet. Some are good some just so so, but I hope it helps people. Just look up ''A Recto'', and I hope to have up a A Champ and A Mesabass and A Hartke soon. BTW I need reading glasses when I type, lol. Great vid bro.
Yes after one week of testing and comparing with the kemper, ToneX is clearly better. Love adding some balls to my chugs with the depth knob!!! It’s simply very accurate. Now I hope kemper will release a software to make the same and make better high gain profiles.
Speaker membrane, impedance, size; Equalizer on a broad range of bands; Pickup Coil output; Strings alloy. That's all you have to worry about when it comes to the tone. From my experimentation it's more about the eq.
12:20 they sound almost identical, but the OG signal seems to have a more prominent high-end fizz. I actually prefer the sound of it using the IR. Whodda guessed it?
@@theedgera2877 I would assume so. But you can't write off what you hear in the room. That why I have a 6505+ specifically for recording and a Fender Master Chorus I run my pedalboard through just for playing in my room. Which sounds amazing, but I'd never use it for recording distorted guitar (I do for clean tones though sometimes). Room tone doesn't have to sound perfect. It only has to sound good to your ear and in the context of your room and placement. These are 2 different things and I feel there should be 2 distinct approaches to them.
Awesome bro! Just got this for my birthday! Haven't started playing much with it yet but it is looking real promising! The tone control, body, and reactivity to this thing is remarkable. Using it alongside Amplitube I image will open entirely new doors!
Thank you for being the first video on RUclips not to lose their mind about ToneX & overlooking things & couldn’t wait to make headlines to rip on kemper & users for even having one lol. Always appreciate & learn a lot from your videos man! I’m stoked for my ToneX & love my Kemper regardless 🤘
So much sound, so little gear. It's hard to lose. There's so many amp sims and multi-effects pedals today that one can question the need of traditional amps and cabinets. The only thing with IK multimedia is that they offer so much you can get option paralysis. Play with purpose. Great video. Thanks.
For the true experimenters out there, check out Guitar ML's Proteus capture based on Alex Wright's Automated-GuitarAmpModelling machine learning project. Same idea using AI to learn how to model distortion and compression of Guitar amps and pedals (no time based effects though). Free and open source plugins using Python and the JUCE audio processing framework....but you don't have to be a programmer to play with it.
It uses GPU processing power to model tones, so you need something like an NVIDIA graphic card. It will shorten dramatically the processing time. Like 40 minutes instead of 5 hours for the advanced mode. Fast mode would only take a minute.
I thought it sounded great for the first release of this idea for the general public, I'm sure they will have big updates in the future. Very neat idea I could picture myself playing around with this thing forever, like I would a new video game becoming a tweak junky and never seeing the outside world ever again.. ..ever ..ever again... ✌
Hey Ola. I wish you would do a video on the factory tones that chug. Yeah, other videos do that, but we have to sit through all the stuff we don't want to hear. When it comes to "will it chug" ..you are the go-to man. When we click on your "Will It Chug" videos, we know that's all we're going to get, a tutorial demonstration on _...."Will it chug-ug-ug-ug?"_
Totally agree. It sounded dry and missed a lot of overtones. Maybe it sounded different out live in his studio but yeah, that live rig sounds really rich and creamy. I would never use the captured tone :S
That phasey tone at 8:38 sounds quite interesting, actually :P Hmm... They are indeed quite close. The two most glaring differences I could hear, are that 1) the ToneX model lacks a bit in the high end (the "fizz" area), and 2) its low end also isn't as big as the real thing... plus, to my ears the chugs come out sounding clippy in some sort of way...
Just FYI guys, if you have an NVIDIA video card, (more vram the better) it will process even the advanced so much faster. I have a 1060 few with 6gb of VRAM, 16g of RAM, i510k and I can do 'advanced' training processes in a little over 30 min, regular on like 15min. Oh and BTW, it HAS to be NVIDIA, apparently the ai uses Cuda cores.
Hello Ola! It´s about time you take that beautifull Ibanez Universe behind you for a spin and make a video? Havent seen you played it yet🤘😎 Cheers from Norway🥳
Tone Models on the right aren't separately paid, if you click on them you just see if you own them in your package SE, regular, or MAX, they are all there.
I bought Tonex along with the AXE I/O as a pre-order specifically to capture all valve amps I have collected over the years that are pretty much sat in storage now because I don't have the space to keep them accessible. I have been using advanced learning and I find it takes me about 40 minutes for that stage of the process, which seems to be a lot quicker than most people are reporting. I don't have a particularly powerful computer set up, just a 2022 M1 MacBook pro with 4gb ram. Looking at the system metrics it's not massively heavy on either CPU or memory.
I guess th biggest difference in sound comes from the slight differences in input impedances. The tonex sim sounds a bit flat and dull. But I do think that it is bypassable by accurately setting up all input gain/impedances...
...close but still far from the mojo of the source. I haven't had luck capturing my Ceriatone Molecular 50W head through a Suhr Reactive Load. I think ToneX has a hard time dealing with direct captures and/or high negative feedback amps (the Molecular has a Girth - neg. feedback - dial & I usually play with it at about 7 on the dial). Hope it improves with time; still happy with my Fractal & Neural plugins.
Hi guys, if there are some nerds who would like to know more about this kind of amp "modelling" I posted a few links in my community section. This technology is quite interessting overall. As a physicist to earn a living we use deep learning in a radiotherapie department to contour the organs at risk in the CT scans.
I used a cheap Berhinger Passive DI Box to capture my amp and it worked just as good as the 300 Canadian box that IK sells!! In fact it worked better than my Active exspensive Radial DI box! I read somewhere that you are supposed to use a Passive Di for better results and it is in fact true!
Sounds to me like comparing the frequency response of the capture and the original, then eq'ing the capture to match might help. The capture seems to lose low end and some hi end.
For sure. Using the Fabfilter Pro-Q 3 match EQ function would work wonders on this. I've done that with my Kemper to match other recorded tones and it works really well. The EQ graph looks crazy but the tones sound alike.
Captures are way quicker if you have an Nvidia GPU (yes it can use the graphics card to process). I'm hoping they add Radeon support or at least let you queue model creation. I've bought it but its too slow with ny current PC, it barely integrates with amplitube and the process of downloading others models via a webpage sucks badly (though the ability to swap the captured cab for at5's IR's is impressive). There's promise but I'm setting it aside for a few patch versions.
Well the profile sounded darker, less definition in the lows but was a good copy. I read hear and there that it gives better result by profiling only the amp out , no miking then go with an IR. Thats just for now, its still kinda beta version. Gotta try your profile asap.
Not all profiles sound great, but some of them sound really freakin good. Definitely better than Overloud's rig player profiles! Love the fact that everyone can make their own profiles and share them. Amplitube has really started to lag behind all other amp sims that are available now, but TONEX makes Amplitube a serious contender again. Btw: it sounds like the captured tone has slightly less gain, but otherwise it's incredibly close. Both Kemper and Neural DSP QC also seems to capture the profiles with slightly less gain so I guess that's something you need to fine tune yourself with all of these profilers.
Ik Multimedia recommends Nvidia GT Seiries for videocard..I just did adavanced capture and training only took 35 minutes. My system is based on an intel i9-9900 chip with 32 gigs of ram and grapics card is a Nvidia GT1650 .
Looks a lot like the free GuitarML stuff, made user friendly for capture. But I love NeuralPi for being able to use the capture as a plugin for the DAW AND as a PEDAL, with a bit of a AI deepdive. Great stuff nevertheless.
Its awesome to see folks like GuitarML and Steve Ack doing free versions of the same thing. I hope they both keep developing and make very easy step-by-step, hand-holding versions for the plebs (like me!) so we can use their software as well. Competition is very good in this space. I hope folks like Neural DSP and Overloud and STL Tones decide to release their versions of this software as well!
@@jasonzdora I wasn't aware of Steve Ack, thanks for pointing out his work, very interesting. His code for the vst3s might cross-compile for RaspberryPi and Elkos, too, with little effort. These projects leave more options for more manipulations on the target sound as you work with regular wav files, like layering amps, comps etc, testing specialized source material, for different instruments or even vocals, using different algorithms/model types and in-depth control.
P.S.: Both guys use lstm models. There is a good chance to make these even compatible. I like that steve even uses 48khz sr, I very much prefer this to 44.1.
For others reading this: Steve Ack's system is called "NAM" (neural amp modeler). Its an open source system that requires a lot of computer know-how (python scripting, other stuff I dont know how to do or even explain) but he can train his system with just a normal audio interface, normal guitar plugged into the DI input and playing random notes (strum up and down the frets with different strength, chugging, tremelo, big open notes, etc) to train NAM to make a totally portable VST3 that can be used anywhere else on any computer without any need for anything else. Pretty awesome. If he can develop it into a step-by-step GUI for idiots (like myself) then he is going to steal the show.
I bought it -- doesn't add the wooly low-mid thing Kemper does (to my ears). I have a fractal FM3, too, but even with that I end up with parametric EQs with all sorts of surgical cuts. At the end of the day, I've noticed I don't have to EQ as much with TONEX; just a low and high pass most of the time and the usual dip at 500-700hz.
the high end difference is bizarre. in both cases you're hearing the mic'ed/recorded sound, ie. the analysis hears the correct high end, but somehow filters it out. I can't think of a good reason for that, and would consider it a bug (unless the machine learning simply can't generate it, but that would seem odd). EDIT. actually I thought of something - it seems the machine learning doesn't model the EQ, but you add a specific type to it afterwards right? if so it could be the added EQ section that is causing the high-freq rolloff. but yeah it would have been great to hear the advanced model, especially as you'd already made one.
I kinda don't see the point in this. Even if you have your "favourite sounding amp" there are effectively infinite different tones so there is always a better sound out there. If you're going to go to the effort of capturing your rig to get something close you may as well just play with a plugin and dial in "another favourite sound". It doesn't take very long to mess around with Guitar Rig or whatever and find something that sounds better than what I was just using. And if you really have to have your special sound then just record the actual rig, in the end it won't matter because after recording it'll play through different headphones/speakers/environment and sound totally different anyway.
Hey ola I'm watching your tonex video and I just heard the new Lamb Of God song to the Grave I was hoping you could talk more about the album if there's anything worth talking about, love your show and can't wait for your new guitar to come out
Maybe I'm missing something, but Positive Grid already did this years ago with their Bias Amp stuff, and it doesn't require special hardware or take 20 minutes to 5 hours, more like a few minutes. Actually, I swear Ola used the Bias Amp in a test against a Line 6, AxeFX, and Kemper a while back as well
How does it work with removing the tone of the guitar you used? The modeling seems like a cool idea, but the guitar used during the process seems to be part of the sound capture too right?
Ola, thanks for insight into this!! Somehow I did skip this software even I use AXE/IO interface with amplitube5. Well, I tried capturing my Carvin Legacy combo with shure SM57. It captured it I would say on 100%. I can't believe that it was that simple! Also I did compute with maximum setting and on old 16cpu (32thr) AMD threadripper system it took ~ 20min to do the full calc. Medium setting took less than 5min to finish just so everyone knows that it can take a bit less than 5 hours. :D I did not try the fast one, but I think it should take less than 2minutes.
@@CoveredByKyle_ Hey, if you have the TONEX plugin you could try my capture. "Legacy LEAD" is the name. If I will have some time to spare I will do some tone examples soon. But your guitars pickups will make big impact on the tone too so the example will be subjective.
What do you guys think? Disregarding the change in level how do they compare? Also the Capture is available in Tonex now.
The Fast capture was definitely underwhelming... I'm curious about the 5 hour version though
Option paralysis
Lacks a little presence on the profile, if you can ad it, it's the same. But the question it's : how does it feel when playing ? Do you feel a big difference in dynamics and sensitivity ?
@@Fanafranky Same. Not that I don't trust Ola but I'd really like to hear the difference.
Hey Ola - regarding the time to create the captures it is relative to what hardware you using to do it - your computer and it’s graphics card apparently - Jason Sadites swapped graphics cards on his machine to an nvidia specified and cut the capture times down to sub 10 minutes for the fast and normal and I believe 27 mins for the advanced super macho capture - I assume you were using your laptop - sounds like for faster captures you need a beefier computer rig. Probably makes sense if your doing a ton of captures for resale online or something.
I have a bunch of great amps in my home studio and got really excited when the new tone x came out with the ability to capture my own amps. The very first time I tried it was with my own radial reamp box out of my interface and I was modeling my Marshall slp 100 with a 4x12 Marshall cab filled with g12h celestions and a royer 121 mic. I went through the entire capture process and chose the "regular" training time. The end result sounded really good but it sounded like my plexi with a pillow over the speaker cab. I then purchased the tone x box itself and ran through the process again using the tone x box to do the re amping work and the box made it sound a touch better but again it lacked the high end clarity of my actual amp mic'ed up. I then starting experimenting around, ditched the ribbon mics and used an sm57 on my Marshall jcm900 and whammmmm. The final capture sounded absolutely identical to my mic'ed up amp. I mean identical and this was using the normal training time. I found the longest training time to only take 1 hour on my iMac but didnt noticed enough of a difference in the final captures over the 23 minutes the "regular" training time took. The key seems to be using a mic that isn't overly dark in the first place. When using the sm57, the average frequency response seems to allow the capture to process in a way that doesn't allow it to become to boomy or filled with low end. I went back and tried my plexi with the royer 121/sm57 combo and leaned the capture mix of the two mics towards the 57 and whammm, the plexi sounded spot on to my mic'ed up tracks going straight into the interface. I now am truly impressed with the tone x capturing. The very best results are making sure you have good high end on the amp settings itself. Meaning, my plexi, tends to be shrill in the high end if I don't back the treble off while playing but when capturing the amp, be sure to allow plenty of this high end through so it can capture the most honest all around tone for your final capture. Just my 2 cents to help others!
very good informative post. thanks!
This is good info. Thank you
There's a reason the SM57/58 have been industry standards since the dawn of time. Great reply, thanks for the info!
Thanks for the info. Will def save me time when I start capturing my stuff.
it's amazing the fact that +10 years after the kemper release we still comparing the tone cloning capabilities vs. him .I mean,10 years it's a lot in terms of technology advances
Yeah and kemper still does it better i am afraid
Kemper is the GOAT to be beaten
@@karolkozak64 Nope lost in low and hi gain comparisons for some time.
Ahhhm I don't think the kemper is nothing special in terms of technology, its brilliance is that it is efficient for the hardware and it established a different mode of working and sharing sound. However it has a lot of sound flaws IF you care so much about that. I mean some of the stuff that bug me from the kemper other types of players (metal vs fusion vs blues etc) would disregard, won't notice or even come in contact with. And in the mix is good enough for many top producers. But this IK multimedia absolutely wins on cost, performance and the ability to just fucking plug it in a daw.
I never liked the kemper. When working as FOH mixer, many guitar guys comming at concerts with that box and always sounding fake in a live situation. It was funny hearing them "but that was the vintage Marshloow model made by Merschuga's guitarrist... it can't be bad". LoL
I think its important to say, that there is a EQ section in the top left corner where you can set up the freqency range for bass, mid and treble. It helps a lot^^ And remember, mesas are really hard to capture. I think the advanced mode will give a bit more sparkle to the tone. On other systems these mode just need 1,5 hours. The Capture got the charackter of the real deal but perhaps a bit tooo much gain =P
It would be interesting to compare the eq profiles of the model vs original sound, and how they differ depending on the training time
To me it sounded boxier, darker than the real thing. I would love to hear the 5 hour capture compared to the real amp
I feel this is a desperate attempt to appeal to what you prefer. They sound nearly the same, I'd argue the capture has better tone lol
@@jmike2039 It has better tone, but it is more static.
@@robertmazurowski5974 what do you mean by static?
It is hard to describe. It is less open and sounds a little bit direct. (Layer on top of direct sound).
@@robertmazurowski5974 huh interesting.
Are you sure you aren't imagining this because of some sort of bias? I mean, would you be able to tell apart a capture and a real amp in a blind test? I don't think I would.
Being that accurate translating the rig tone I believe it is worth the wait on the processing time. Better to spend some extra time getting the perfect tone before and doing the job just once. Great review btw. Thanks!
In machine learning longer training time doesn't essentially mean a better model (over fitting, less generalisation).
I wonder how exactly this was optimized here...
@@CraftingCake Hi. I understand it could be not proportionally better, but shouldn't it be better? You mean then that a longer training time could result in zero improvement or worse, a negative improvement?? I'm not into machine learning, but made my comment based purely on common sense
Great vid, you don"t know me but you mentioned my captures up for free on their Tonenet. Some are good some just so so, but I hope it helps people. Just look up ''A Recto'', and I hope to have up a A Champ and A Mesabass and A Hartke soon. BTW I need reading glasses when I type, lol. Great vid bro.
Yes after one week of testing and comparing with the kemper, ToneX is clearly better. Love adding some balls to my chugs with the depth knob!!! It’s simply very accurate. Now I hope kemper will release a software to make the same and make better high gain profiles.
Speaker membrane, impedance, size;
Equalizer on a broad range of bands;
Pickup Coil output;
Strings alloy.
That's all you have to worry about when it comes to the tone. From my experimentation it's more about the eq.
Nice program overall, that's pretty exciting! great vid ola, killin' it as always!
12:20 they sound almost identical, but the OG signal seems to have a more prominent high-end fizz. I actually prefer the sound of it using the IR. Whodda guessed it?
@@theedgera2877 I would assume so. But you can't write off what you hear in the room. That why I have a 6505+ specifically for recording and a Fender Master Chorus I run my pedalboard through just for playing in my room. Which sounds amazing, but I'd never use it for recording distorted guitar (I do for clean tones though sometimes). Room tone doesn't have to sound perfect. It only has to sound good to your ear and in the context of your room and placement.
These are 2 different things and I feel there should be 2 distinct approaches to them.
Now I wanna hear the 5 hour model
Awesome bro! Just got this for my birthday! Haven't started playing much with it yet but it is looking real promising! The tone control, body, and reactivity to this thing is remarkable. Using it alongside Amplitube I image will open entirely new doors!
Thank you for being the first video on RUclips not to lose their mind about ToneX & overlooking things & couldn’t wait to make headlines to rip on kemper & users for even having one lol.
Always appreciate & learn a lot from your videos man! I’m stoked for my ToneX & love my Kemper regardless 🤘
So much sound, so little gear. It's hard to lose. There's so many amp sims and multi-effects pedals today that one can question the need of traditional amps and cabinets. The only thing with IK multimedia is that they offer so much you can get option paralysis. Play with purpose. Great video. Thanks.
For the true experimenters out there, check out Guitar ML's Proteus capture based on Alex Wright's
Automated-GuitarAmpModelling machine learning project. Same idea using AI to learn how to model distortion and compression of Guitar amps and pedals (no time based effects though). Free and open source plugins using Python and the JUCE audio processing framework....but you don't have to be a programmer to play with it.
Love the tone of the rig and how well it was able to capture the tone
It's all fun and games until ToneNet becomes self aware..............
And then it creates a brand new hi gain digital sim to conquer all vst amp sims. 🙃🙃
Self aware? Didn't get it?
@@mylogify A Terminator reference ;)
@@dustinsmith8658 How?
@@shaharbar Which Terminator scene haha?
I have no idea what's going on but I love it.
It uses GPU processing power to model tones, so you need something like an NVIDIA graphic card. It will shorten dramatically the processing time. Like 40 minutes instead of 5 hours for the advanced mode. Fast mode would only take a minute.
I thought it sounded great for the first release of this idea for the general public, I'm sure they will have big updates in the future. Very neat idea I could picture myself playing around with this thing forever, like I would a new video game becoming a tweak junky and never seeing the outside world ever again.. ..ever ..ever again... ✌
Ha! I've been waiting for this since you been born! Thanks!
The one video I was waiting for! Awesome review as always!
eh? The guitar goes in the hole ? Ola you seem frustrated 🥴 and this is a no go for me , I will hold onto my amps and quad cortex
"I'm not gonna wait for 5 hours for a new profile, ya know?"
Me remembering taking plane trips just to visit a vintage amp seller.
Hey Ola. I wish you would do a video on the factory tones that chug. Yeah, other videos do that, but we have to sit through all the stuff we don't want to hear. When it comes to "will it chug" ..you are the go-to man. When we click on your "Will It Chug" videos, we know that's all we're going to get, a tutorial demonstration on _...."Will it chug-ug-ug-ug?"_
🍓
Wirklich unheimlich ... 😂
The real thing is more alive and just bigger sounding, more bass, more tones!
I literally just looked up portable amp capture yesterday and found this. Ola knows the algorithm better than Google.
I've been asking for ages about some VST Kemper profile player and I glad competitors picked up the idea that was in the air
I thought that sounded way different and not good lol. Hearing Ola say that sounds pretty good and accurate is blowing my mind a little bit.
Sounds like muffled high end a little to me
I think it sounds very close if you disregard the difference in level. Definitely not something you would hear in a mix.
Totally agree. It sounded dry and missed a lot of overtones. Maybe it sounded different out live in his studio but yeah, that live rig sounds really rich and creamy. I would never use the captured tone :S
I guess you haven’t figured out that tone is subjective yet
@@tommilitello198 I guess you haven't figured out what "I thought" means yet.
Interesting. The capture isn't identical but very close. The real rig has an openness / richness to it that the capture doesn't have.
That phasey tone at 8:38 sounds quite interesting, actually :P
Hmm... They are indeed quite close. The two most glaring differences I could hear, are that 1) the ToneX model lacks a bit in the high end (the "fizz" area), and 2) its low end also isn't as big as the real thing... plus, to my ears the chugs come out sounding clippy in some sort of way...
Just FYI guys, if you have an NVIDIA video card, (more vram the better) it will process even the advanced so much faster. I have a 1060 few with 6gb of VRAM, 16g of RAM, i510k and I can do 'advanced' training processes in a little over 30 min, regular on like 15min.
Oh and BTW, it HAS to be NVIDIA, apparently the ai uses Cuda cores.
“Ah it hertz” -the correct response to a frequency sweep hitting your ears
Hello Ola! It´s about time you take that beautifull Ibanez Universe behind you for a spin and make a video? Havent seen you played it yet🤘😎 Cheers from Norway🥳
Almost a good a question as 'Will IK ever stop sending me 5 emails a day about this?'
I love your honesty. Five effing hours!! TONEX... Potential
Tone Models on the right aren't separately paid, if you click on them you just see if you own them in your package SE, regular, or MAX, they are all there.
I think the modeling has greatly improved to catch up with NAM plugin, but not quite sure
The Fast capture has high-mid and high details missing. I'd love to hear the advanced capture.
I bought Tonex along with the AXE I/O as a pre-order specifically to capture all valve amps I have collected over the years that are pretty much sat in storage now because I don't have the space to keep them accessible. I have been using advanced learning and I find it takes me about 40 minutes for that stage of the process, which seems to be a lot quicker than most people are reporting. I don't have a particularly powerful computer set up, just a 2022 M1 MacBook pro with 4gb ram. Looking at the system metrics it's not massively heavy on either CPU or memory.
Because TensorFlow (engine that handles learning underground) can use M1, otherwise you should have NVidia GPUs.
I guess th biggest difference in sound comes from the slight differences in input impedances. The tonex sim sounds a bit flat and dull. But I do think that it is bypassable by accurately setting up all input gain/impedances...
Ola you need to find this Paul Steven Cocks fella and tell him you tried his tone..
Now that they release the ToneX Pedal, please please, make a proper TubeScreamer-Badlander-MesaOversized4x12 profile.
Pros : Captures your tone
Cons : no signal when using Free The Tone pedals
Dude no I already have an unhealthy addiction to buying plugins. Im gonna need a sponsor soon.
...close but still far from the mojo of the source. I haven't had luck capturing my Ceriatone Molecular 50W head through a Suhr Reactive Load. I think ToneX has a hard time dealing with direct captures and/or high negative feedback amps (the Molecular has a Girth - neg. feedback - dial & I usually play with it at about 7 on the dial). Hope it improves with time; still happy with my Fractal & Neural plugins.
It's amazing how well these companies are using capturing if you ask me. My simple minded way of seeing it, is they're just advanced guitar pedals.
This is like social media for Amps.
Hi guys, if there are some nerds who would like to know more about this kind of amp "modelling" I posted a few links in my community section.
This technology is quite interessting overall. As a physicist to earn a living we use deep learning in a radiotherapie department to contour the organs at risk in the CT scans.
Besides: Thank you, Ola for your presentation. Very cool!
Hey Happy Belated Birthday Ola! Hope it was awesome.
Please Ola make a Century 200 patch for us mere mortals!
the capture is a little darker and has less bite.
I used a cheap Berhinger Passive DI Box to capture my amp and it worked just as good as the 300 Canadian box that IK sells!! In fact it worked better than my Active exspensive Radial DI box! I read somewhere that you are supposed to use a Passive Di for better results and it is in fact true!
Sounds to me like comparing the frequency response of the capture and the original, then eq'ing the capture to match might help. The capture seems to lose low end and some hi end.
For sure. Using the Fabfilter Pro-Q 3 match EQ function would work wonders on this. I've done that with my Kemper to match other recorded tones and it works really well. The EQ graph looks crazy but the tones sound alike.
@@SomebodyPickaName It's weird how their software misses in this aspect.
That sounds fucking insane....I can't tell the difference honestly.
Captures are way quicker if you have an Nvidia GPU (yes it can use the graphics card to process). I'm hoping they add Radeon support or at least let you queue model creation.
I've bought it but its too slow with ny current PC, it barely integrates with amplitube and the process of downloading others models via a webpage sucks badly (though the ability to swap the captured cab for at5's IR's is impressive). There's promise but I'm setting it aside for a few patch versions.
Don't hold your breath on Radeon support. AMD isn't even worth mentioning when it comes to AI, CUDA is the uncontested king of machine learning.
@@17Codiferus Yeah my PC is due for an upgrade and this might swing me toward nvidia (depending how greedy they get)
This is totally game changing. wow!
Hi Ola, please try new Mercuriall Dual Rectofire plugin, it seems like the best rectifier modeler. It would be very interesting to know your opinion
I'm getting this thing, really awesome product
Well the profile sounded darker, less definition in the lows but was a good copy.
I read hear and there that it gives better result by profiling only the amp out , no miking then go with an IR. Thats just for now, its still kinda beta version.
Gotta try your profile asap.
"Let's just push NEXT!"
Ola the capturer. Need to check this Tonex out.
The ToneX capture sounded just a little bit tighter than the real thing.
Not all profiles sound great, but some of them sound really freakin good. Definitely better than Overloud's rig player profiles! Love the fact that everyone can make their own profiles and share them. Amplitube has really started to lag behind all other amp sims that are available now, but TONEX makes Amplitube a serious contender again.
Btw: it sounds like the captured tone has slightly less gain, but otherwise it's incredibly close. Both Kemper and Neural DSP QC also seems to capture the profiles with slightly less gain so I guess that's something you need to fine tune yourself with all of these profilers.
Ik Multimedia recommends Nvidia GT Seiries for videocard..I just did adavanced capture and training only took 35 minutes. My system is based on an intel i9-9900 chip with 32 gigs of ram and grapics card is a Nvidia GT1650 .
By the way, if you have a GPU, it is considerably faster. With a nvidia gtx 1660 super here the advanced training takes aroud 25 minutes.
Whole new series... "Will it Capture?"
The AI training time is entirely dependent on you GPU. I have a 3080ti and the advanced training takes 20 Min.
i really wanna see the demo of NAM
Looks a lot like the free GuitarML stuff, made user friendly for capture. But I love NeuralPi for being able to use the capture as a plugin for the DAW AND as a PEDAL, with a bit of a AI deepdive. Great stuff nevertheless.
Its awesome to see folks like GuitarML and Steve Ack doing free versions of the same thing. I hope they both keep developing and make very easy step-by-step, hand-holding versions for the plebs (like me!) so we can use their software as well. Competition is very good in this space. I hope folks like Neural DSP and Overloud and STL Tones decide to release their versions of this software as well!
@@jasonzdora I wasn't aware of Steve Ack, thanks for pointing out his work, very interesting. His code for the vst3s might cross-compile for RaspberryPi and Elkos, too, with little effort. These projects leave more options for more manipulations on the target sound as you work with regular wav files, like layering amps, comps etc, testing specialized source material, for different instruments or even vocals, using different algorithms/model types and in-depth control.
P.S.: Both guys use lstm models. There is a good chance to make these even compatible. I like that steve even uses 48khz sr, I very much prefer this to 44.1.
@@imcrazedandconfused Its pretty exciting to see where this is all going to end up!
For others reading this: Steve Ack's system is called "NAM" (neural amp modeler). Its an open source system that requires a lot of computer know-how (python scripting, other stuff I dont know how to do or even explain) but he can train his system with just a normal audio interface, normal guitar plugged into the DI input and playing random notes (strum up and down the frets with different strength, chugging, tremelo, big open notes, etc) to train NAM to make a totally portable VST3 that can be used anywhere else on any computer without any need for anything else. Pretty awesome. If he can develop it into a step-by-step GUI for idiots (like myself) then he is going to steal the show.
I bought it -- doesn't add the wooly low-mid thing Kemper does (to my ears). I have a fractal FM3, too, but even with that I end up with parametric EQs with all sorts of surgical cuts. At the end of the day, I've noticed I don't have to EQ as much with TONEX; just a low and high pass most of the time and the usual dip at 500-700hz.
Tonex won't open in my Studio One. IK told me two weeks ago that a fix would be 'very soon'.
Laney Cub-Supertop for a Will it chug!
the GE300 has an excellent tone capture process and can capture pedals, amps and guitars.
Mooer is crap that won't last 3 winters. Enjoy it while it still works. With Tonex, you can at least get a new laptop.
Holy crap, it actualy sounds better, less noise! :)
impressive ! nice demo
That is really impressive, I’ve liked IK, although their software has always seemed bloated. But this looks pretty interesting.
that's fair. they have like 7 different JCM800's by themselves. maybe 4 Rectifiers. it really gives one choice anxiety.
They should make an option to go with the fast setting and then letting the advanced setting render overnight.
The Tonex is definitely darker.
at 13:45.....oh shhhhhhiiiiittttt
the high end difference is bizarre. in both cases you're hearing the mic'ed/recorded sound, ie. the analysis hears the correct high end, but somehow filters it out. I can't think of a good reason for that, and would consider it a bug (unless the machine learning simply can't generate it, but that would seem odd). EDIT. actually I thought of something - it seems the machine learning doesn't model the EQ, but you add a specific type to it afterwards right? if so it could be the added EQ section that is causing the high-freq rolloff. but yeah it would have been great to hear the advanced model, especially as you'd already made one.
I kinda don't see the point in this. Even if you have your "favourite sounding amp" there are effectively infinite different tones so there is always a better sound out there. If you're going to go to the effort of capturing your rig to get something close you may as well just play with a plugin and dial in "another favourite sound". It doesn't take very long to mess around with Guitar Rig or whatever and find something that sounds better than what I was just using.
And if you really have to have your special sound then just record the actual rig, in the end it won't matter because after recording it'll play through different headphones/speakers/environment and sound totally different anyway.
I used ATH-M50X and the differences are quite huge.
I can’t believe he missed the paulstevenscox joke.
Of course it sounded good
To me it sounded like free guitar plugin.
Hey ola I'm watching your tonex video and I just heard the new Lamb Of God song to the Grave I was hoping you could talk more about the album if there's anything worth talking about, love your show and can't wait for your new guitar to come out
Maybe I'll double the possible question but... the time of "Accuracy Level" depends on the PC computing power ? or it's done "somewhere" else.
has anyone analysed the spectrum of the differential signal?
L²-norm? That'd be neat.
Sounds great but do we need the 'Capture' machine to capture the tone/ Can we do it with just the Tonex Pedal?
Thanks was wondering about tonex.
Maybe I'm missing something, but Positive Grid already did this years ago with their Bias Amp stuff, and it doesn't require special hardware or take 20 minutes to 5 hours, more like a few minutes. Actually, I swear Ola used the Bias Amp in a test against a Line 6, AxeFX, and Kemper a while back as well
Yeah its really a copy of the bias ecosystem. But the tones are better with tonex imo
It's finally out!!
It's pretty close but it has some weird strange overtone especially up high on the model.
Just before this video downloded a free version ov this vts. So thanks for the great video. Be safe and skilful. Pinge
How does it work with removing the tone of the guitar you used? The modeling seems like a cool idea, but the guitar used during the process seems to be part of the sound capture too right?
Ola, thanks for insight into this!! Somehow I did skip this software even I use AXE/IO interface with amplitube5. Well, I tried capturing my Carvin Legacy combo with shure SM57. It captured it I would say on 100%. I can't believe that it was that simple! Also I did compute with maximum setting and on old 16cpu (32thr) AMD threadripper system it took ~ 20min to do the full calc. Medium setting took less than 5min to finish just so everyone knows that it can take a bit less than 5 hours. :D I did not try the fast one, but I think it should take less than 2minutes.
The legacy is an extremely hard amp to replicate- any chance you could send some tone examples? Or, something like that :)
@@CoveredByKyle_ Hey, if you have the TONEX plugin you could try my capture. "Legacy LEAD" is the name. If I will have some time to spare I will do some tone examples soon. But your guitars pickups will make big impact on the tone too so the example will be subjective.
I've been playing through an Eleven Rack for years. Wondering if this would be a good replacement for that. I'm talking about the new ToneX Pedal.